Introduction

If you think credit scoring as a common case of machine learning application, yes it is true. We knew that this credit scoring model is really useful to predict the credit worthiness of someone. With this model lender can decide to approve the credit application faster and more effective. It is advantage for both of lender and debtor.

image1. credit report

On the internet, there are many sources explaining how to build the credit scoring model as a general modeling case. They often use germancredit as an example and it looks very easy. However, as someone who ever work in credit bureau, I think the problem was not that simple. There are a problem when we would like to start developing a credit scoring model. The problem that might happened in credit scoring case as follow:

Data Sources

In order to build a behaviour credit scoring, we can use almost every data such as telecom, electroicty, tax, water payment and obviously credit data. As long as we

Identity matching

You can also embed plots, for example:

Note that the echo = FALSE parameter was added to the code chunk to prevent printing of the R code that generated the plot.

Bad & Good Definition

Period selection

Feature engineering

Univariate analysis

Descriptive analysis

Modeling

Evaluation

Deployment